Research On In-air Chinese Signature Authentication Based On Multi-sensor Fusion

The rapid development of mobile payment makes the security of payment become an urgent problem to be solved.The authentication of user identity is the core of the problem,which is the first pass for users to hold their personal assets.The insecurity of traditional password authentication makes the in-air signature authentication technology based on the sensor of smartphones develop rapidly.However,the current in-air signature authentication technology cannot guarantee the accuracy of authentication and the speed of signature matching at the same time.In view of the existing shortcomings,the signature authentication method based on time function is studied.Based on the structural characteristics of Chinese characters,this paper proposes a signature verification method that segmentation first then matches.In the process of identity authentication,the axis will merged first and then the fusion between sensors is performed.Finally,high-precision,high-efficiency identity authentication is achieved.The main work of this paper is as follows:Acquires in-air Chinese signature data by using smart phones which accelerometer and gyroscopes.And the data must be interpolated and resampled first because of the non-uniform sampling of the mobile phone.Then take a differential threshold method to intercept valid signature data.Finally,the signal will be filtered,and the length of the signature will be adjusted to 500 points.In order to avoid the problem of low matching rate and slow matching speed caused by traditional global alignment algorithm,we propose a method of extreme point matching,segmentation and then measure the signature distance.In order to solve the problem of extreme point polarity mismatch caused by the traditional GSA algorithm,we improve the GSA algorithm and constraint algorithm matching path,so that the extreme points of the same polarity can be correctly matched.Then we propose to use the DTW matching method to select the one-to-many matching pairs of extreme points,and obtain the best extreme point matching pairs in the sequence.and complete the sequence segment alignment and similarity measurement.Finally,the sequence alignment and similarity measure are completed.The authentication effect of each axis of the two sensors has been effectively improved when use the segment matching algorithm,but there are differences in the authentication effects of different axes.Compare and analyze the certification effects under multi-axis fusion and remove the redundant axis.Finally,the information of the two sensors is integrated.Distance results obtained by the two sensors are transferred to confidence.Then the voting method and the weighted average method are applied to fuse the information on the decision layer.The database of the experiment is a self built data set,including a total of 2000(50*40)in-air Chinese signatures of 50 users.The experimental results show that the method that each axes of the sensor segmented first and measured similarity,then multi-axis fusion,and finally the fusion of the two sensors can improve the system efficiency and certification accuracy.And a result of FRR: 1.88% and FAR: 1.9% is obtained.